[FieldTrip] how to find significant time point and electrode clusters after permutation test in fieldtrip

Schoffelen, J.M. (Jan Mathijs) janmathijs.schoffelen at donders.ru.nl
Fri Aug 26 10:05:26 CEST 2022

Hi Jaewon,

I suggest that you check the e-mail archive (using Google or so). Your question pops up occasionally, and has been answered before, e.g. at https://mailman.science.ru.nl/pipermail/fieldtrip/2019-May/039144.html

As to why (and in which way) the outputs are different, I don’t know. Your screenshots did not come through in your message, probably because they were too big in size. Are you sure that you used the exact same cfg in the call to ft_timelockstatistics? If so, could it be that in one of the analyses the a priori cluster thresholding scheme did not yield any clusters in the observed data?

Good luck and best wishes,

On 17 Aug 2022, at 18:39, 오재원 via fieldtrip <fieldtrip at science.ru.nl<mailto:fieldtrip at science.ru.nl>> wrote:

Hello, my name is Jaewon.
Currently I am trying cluster-based permutation tests with EEG/ERP data.
I have finished writing the code script for the permutation tests (until ft_timelockstatistics) and my matlab script did not give an error.
However, I could not find where I can see significant time point and electrode clusters.

I conducted two permutation tests : one for 0-600ms post stimuli onsets (to see N400 effect) and the other for 600-1000ms (to see P600 effect).
I made 'stat' variable as a result of using 'ft_timelockstatistics' function.
Attached files are Matlab screens of fields resulted from permutation tests.

In the case of 0-600ms post stimulus onset, 'posclusters' seemed to show some statistical stuffs like probability, standard deviation, CI range, etc. Nonetheless, what I was looking for was significant time domain and electrodes as well as corresponding p-values. But I could not find those here. In the case of 600-1000ms post stimulus onset, the 'stat' variable did not create some fields such as 'prob', 'posclusters', 'posclusterslabelmat', and so on. Matlab scripts for the two analyses are identical except for cfg.latency. (For 600-1000ms analysis, 'cfg.latency = [0.6 1.2]' was used instead of 'cfg.latency = [0 0.6]') So I also did not understand why the results are different while the scripts are the same.

To summarize, I am wondering how and where I can find significant time point and electrode clusters and corresponding p-vlalues after permutation test in fieldtrip.
Additionally, I would like to know why the results of 0-600ms analysis and of 600-1000ms analysis differ in spite of the same scripts.
Thank you for reading my email and any advice or help would be greatly appreciated.

All the best,

fieldtrip mailing list

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20220826/0f95d930/attachment.htm>

More information about the fieldtrip mailing list